JP2002135779A - Re-encoding apparatus of image signal - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a re-encoding apparatus of image signal capable of enhancing efficiency of encoding by detecting picture types from image signals when re-encoding the image signals subjected to an encoding processing. SOLUTION: In the apparatus, to perform re-encoding processing for decoded image signals subjected to the encoding processing as input image signals, a DCT device 50 to perform discrete cosine transformation to the input image signals, a DCT coefficient counter 51 to calculate feature value of every picture by using DCT coefficients outputted from the device 50, a picture type detector 52 to detect the picture types at the encoding processing in the prestage by using the feature value outputted from the counter 51, an encoding control part 11 to determine encoding parameters at the re-encoding responding to detected results of the detector 52 and an encoding part 12 to perform the re-encoding processing by using the parameters determined at the control part 11.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of re-encoding an image signal once decoded after being encoded.
The present invention relates to an image signal re-encoding device capable of minimizing encoding deterioration in re-encoding even if an encoding parameter in the preceding encoding is not obtained.

[0002]

2. Description of the Related Art Various encoding methods for reducing the amount of information of a digital image signal have been proposed, and a plurality of methods have been established as international standards. Among them, the coding method (for example, I
H. TU-T Recommendation. 261 and H.E. 263, ISO / IE
If an attempt is made to edit an image signal encoded according to C standard MPEG or the like, there are various restrictions in performing the operation in the state of an encoded bit stream. The work is easier if the user edits from the beginning and follows the procedure of re-encoding.

[0005] However, if the encoded image signal is decoded once and then re-encoded, usually significant coding deterioration occurs. In order to avoid this, it is known that re-encoding processing should be performed using the same encoding parameters used in the preceding encoding processing. As the encoding parameter, for example, a “picture type” indicating a prediction method such as intra-frame prediction (I picture) / forward inter-frame prediction (P picture) / bidirectional inter-frame prediction (B picture), and quantization processing are used. And a "motion vector" used for motion compensation prediction. Among them, “picture type” has the greatest effect on the image quality of the re-encoded image and is an important parameter.

[0004] These parameters can be output together with the decoded image signal from a decoder that decodes the coded bit stream into an image signal, but the decoder does not have a parameter output function or is output. It is also assumed that only the decoded image signal is obtained due to lost parameter data. As a method for extracting parameters from a decoded image signal when only such a decoded image signal is obtained, for example, there is a method disclosed in Japanese Patent Laid-Open No. 10-32829.

FIG. 6 is a block diagram showing the configuration of the re-encoding device disclosed in Japanese Patent Laid-Open No. 10-32829.
6, reference numeral 10 denotes a preprocessing unit, 11 denotes an encoding control unit,
Reference numeral 12 denotes an encoding unit. The preprocessing unit 10 includes an intra-frame encoder 20, an SNR calculator 21, and an I-picture detector 22.

Next, the operation will be described. The input image signal 31 input to the re-encoding device is obtained by decoding an encoded bit stream by a decoder (not shown).
It is assumed that the image has been encoded at least once. The intra-frame encoder 20 outputs all the input image signals 3
1 is intra-coded using a fixed quantization step size, and a bit stream 41 is output. The SNR calculator 21 calculates the SN from the encoded bit stream 41.
Calculate the R value and output the SNR value 42. The I picture detector 22 detects the position of the I picture from the value of the SNR value 42. The encoding control unit 11 includes an I picture detector 22
And outputs a control signal 32 for controlling the encoding process of the encoding unit 12 using the result detected by the control unit. Encoding unit 12
Encodes the input image signal 31 based on the control signal 32, and outputs an encoded bit stream 33.

[0007]

According to the conventional example described above, the SNR calculator 21 measures the SNR of an image signal, but the method is not specified. In the device configuration shown in FIG. 6, in order to measure the SNR in the SNR calculator 21, the input bit stream 41 is temporarily decoded to obtain a decoded image signal, and the input image signal 31 is subjected to arithmetic processing ( In order to obtain the SNR, it is necessary to perform a sum of square differences).

Further, since the intra-frame encoder 20 needs to output the bit stream 41, when using an encoding method such as MPEG, at least the DCT
(Discrete cosine transform) unit, quantizer, variable length coding (V
An LC) device is required.

As described above, the conventional re-encoding device includes the intra-frame encoder 20 and the SNR calculator 21 of the pre-processing unit.
However, there is a problem that the device configuration becomes large and the amount of calculation becomes enormous. Although the method of detecting an I picture is specified, the P picture /
It does not specify a method for detecting a B picture.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problem, and it is an object of the present invention to improve coding efficiency by detecting a picture type from an image signal when re-encoding an image signal that has undergone an encoding process. It is an object of the present invention to obtain an image signal re-encoding device capable of performing the following.

[0011]

An image signal re-encoding device according to the present invention uses a decoded image signal that has been subjected to an encoding process as an input image signal and performs a re-encoding process on the image signal. A DCT unit for performing a discrete cosine transform of the input image signal, and a DT output from the DCT unit.
DCT that counts the amount of features for each picture using CT coefficients
A coefficient counter, a picture type detector that detects a picture type in a preceding encoding process using the feature quantity output from the DCT coefficient counter, and a picture type detector based on the detection result of the picture type detector. An encoding control unit for determining an encoding parameter, and an encoding unit for performing a re-encoding process using the encoding parameter determined by the encoding control unit are provided.

Further, the picture type detector detects any one of three picture types as an intra-coded picture, a forward inter-frame predictive coded picture, and a bidirectional inter-frame predictive coded picture as a picture type to be detected. It is characterized by including one or more.

Further, the DCT coefficient counter obtains a sum of absolute values or a sum of squares of DCT coefficients for each frequency domain as a feature quantity to be counted. It is characterized in that the picture type is detected in accordance with the temporal change of the sum.

Further, the picture type detector detects a picture in which the sum of absolute values or sum of squares in a high frequency region is smaller than the preceding and succeeding pictures as an intra-frame coded picture.

Further, the picture type detector detects a picture in which the sum of absolute values or sum of squares in a low frequency region is larger than the preceding and succeeding pictures as an intra-frame coded picture or a forward inter-frame coded picture. It is assumed that.

Further, the DCT coefficient counter obtains, as a feature quantity to be counted, the number of DCT coefficients whose absolute value is larger than a predetermined threshold value or smaller than the predetermined threshold value. , A picture type is detected in accordance with the obtained number.

Further, the picture type detector has a DC
It is characterized in that a picture in which the number of absolute values of the T coefficient is larger than a threshold value is small or a picture in which the absolute value of the T coefficient is smaller than the threshold value is large is detected as an intra-frame coded picture.

Further, the encoding control unit determines an encoding parameter using a picture type detected by the picture type detector.

Further, the encoding control unit determines an encoding parameter using a target generated code amount set according to the picture type detected by the picture type detector. .

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiment 1 FIG. 1 is a schematic block diagram showing a configuration of an image signal re-encoding device according to Embodiment 1 of the present invention. In FIG. 1, preprocessing section 13 according to the first embodiment coefficients DCT unit 50 for performing discrete cosine transform of an input image signal and DCT coefficients output from DCT unit 50 to calculate a feature amount for each picture. D
It comprises a CT coefficient counter 51 and a picture type detector 52 for detecting a picture type in the preceding encoding process using the feature quantity output from the DCT coefficient counter 51, and the other parts are the same as those of the conventional example. , The encoding control unit 11 determines the encoding parameters at the time of re-encoding according to the detection result of the picture type.
2 performs re-encoding processing using the determined encoding parameters.

Next, the operation will be described. Preprocessing unit 1
The input image signal 31 is input to the third DCT unit 50.
The DCT unit 50 performs the same two-dimensional DCT as that performed in normal image coding, and outputs a DCT coefficient 61 composed of 64 frequency components. FIG. 2 shows the frequency characteristics of the 64 DCT coefficients output from the DCT unit 50.

The DCT coefficient counter 51 calculates a sum of absolute values or a sum of squares for each frequency component for each picture, and outputs a power value 62 of each frequency component. Alternatively, an average value may be obtained for each frequency component, and a sum of absolute differences or a sum of square differences from the average value may be obtained.

The picture type detector 52 detects the picture type of each picture. FIG. 3 shows high-frequency components of both the horizontal and vertical components of the DCT count (the hatched portions in FIG. 2).
FIG. 4 shows the power value of the DCT coefficient when the horizontal component is low and the vertical component is high (the horizontal line in FIG. 2), or the DCT coefficient is high when the horizontal component is high and the vertical component is low. It indicates the power value of the low band (the vertical line portion in FIG. 2), and a picture which is an I picture in the preceding encoding process is, for example, a P picture or a B picture as indicated by a circle in the graph of FIG.
Since the high frequency component of the DCT coefficient tends to be poorer than that of the picture, a picture in which the power value 62 of the high frequency component is smaller than that of the preceding and following pictures is detected as an I picture.
That is, a picture in which the difference between the power values of the preceding and following pictures is smaller than a preset value is detected as an I picture.

In the preceding encoding process, pictures that are I pictures or P pictures are, for example, as shown in FIG.
Since the power value of the low frequency component of the DCT coefficient tends to be larger than that of the B picture as shown by the circles in the graph of FIG. Detected as That is, a picture in which the difference between the power values of the preceding and succeeding pictures is greater than a preset
Detect as a picture. For this reason, if FIG. 3 and FIG. 4 are combined, an I picture is detected from FIG. 3. Therefore, except for the position of a circle detected as an I picture from FIG. it can.

The coding control unit 11 sets various coding parameters according to the picture type detected by the picture type detector 52, and controls the coding process of the coding unit 12.

The coding control section 11 controls the coding process to be performed with the same picture type as the picture type detected by the picture type detector 52. Thereby, re-encoding processing with little image quality degradation can be executed. Further, the encoding control unit 11 sets a target code amount according to the detected picture type and controls the encoding unit. This allows
Code amount distribution suitable for the bit rate of the bit stream 33 output by the re-encoding process can be executed.

A plurality of DCTs among 64 frequency components
The low frequency components (for example, unmarked portions 16 in FIG. 2)
It is also effective to preliminarily specify the coefficients as high-frequency components (for example, 16 hatched portions in FIG. 2), and obtain the absolute value sum and the sum of squares of the plurality of coefficients in the DCT coefficient counter 51. In this case, the output power value 6
2 is not 64, but is output by the number of components specified in advance.

When one frame signal is an interlaced signal composed of two field signals, a method of separately encoding the two field signals (called a field structure in the case of MPEG-2) is combined into one. (In the case of MPEG-2, called a frame structure). In the case of the frame structure, two field signals are alternately arranged, so that the size occupied by one block with respect to the screen is almost a square in both horizontal and vertical directions. In the case of the field structure, since it is composed of only one field signal, it has a vertically long rectangular state in which the vertical direction is almost twice as large as the horizontal direction. Therefore, in the case of an interlaced signal, it is known that the correlation in the vertical direction is lower in both the field structure and the frame structure than in the horizontal direction.

Therefore, when the input image signal is an interlaced signal, many signals are generated for the high frequency component in the vertical direction. Therefore, the detection using the low frequency component and the high frequency component in the vertical direction is not performed, and the detection in the horizontal direction is not performed. The picture type is detected using the low-frequency component (for example, the area of the horizontal line in FIG. 2) and the high-frequency component (the area of the hatched part in FIG. 2). This makes it possible to increase the accuracy of detection.

Embodiment 2 FIG. Next, an image signal re-encoding device according to Embodiment 2 will be described. The configuration of the image signal re-encoding device according to Embodiment 2 is the same as that of Embodiment 1 shown in FIG. Here, a threshold value is set in the DCT coefficient counter 51 according to the second embodiment in advance, and the number of DCT coefficients whose absolute value components are larger or smaller than the threshold value for 64 DCT coefficients is calculated. Has been made. Pictures that are I-pictures in the preceding encoding process tend to generate more DCT coefficients having a value of 0 or near 0 than P-pictures and B-pictures. FIG. 5 shows a time transition of the number of DCT coefficients whose absolute value is larger than a threshold value.
The circles indicate positions that were I-pictures in the preceding encoding process. Therefore, the picture type detector 52 detects, as an I picture, a picture in which the number of DCT coefficients larger than the threshold is smaller than that of the preceding and following pictures, or a picture in which the number of DCT coefficients smaller than the threshold is larger.

Further, by combining the first embodiment and the second embodiment, it is possible to further improve the detection accuracy. For example, when FIG. 3 and FIG. 5 are used together, the detection accuracy of the I picture can be further improved. Further, a P picture can be specified and detected using FIGS. 5 and 4. Also, the preprocessing unit 13 may extract a part of the screen and input the input image signal instead of inputting the entire input image signal to the DCT unit 50. The processing amount can be reduced.

In the first and second embodiments, the preprocessing unit 13 and the encoding unit 12 have been described as separate units. However, the DCT unit and the preprocessing unit 13 included in the encoding unit 12 are described. It is also possible to use a common DCT unit. In this case, it is possible to reduce the circuit of the DCT unit by sharing.

The input image signal processed by the encoding unit 12 as an I-picture or the macroblock subjected to intra-frame encoding processing even if processed by the encoding unit 12 as a P-picture or a B-picture. On the other hand, since the same DCT coefficient output from the DCT unit 50 of the preprocessing unit is generated in the encoder 12, the DCT
It is also possible to store the result of the device 50 in a memory or the like and reuse it. In this case, the DCT in the encoder 12
Can be reduced.

Further, in the first embodiment, the description has been given by taking as an example the "picture type" and the "target code amount" as the encoding parameters controlled by the encoding control unit 11. For example, in the case of MPEG encoding, It is also possible to control other encoding parameters such as a "quantization matrix", a "variable-length codeword table selection number", and a "motion vector search range" that can be set for each picture.

[0035]

As described above, according to the present invention, a DCT unit for performing discrete cosine transform of an input image signal,
A DCT coefficient counter that counts a feature amount for each picture using a DCT coefficient output from the T unit, and a picture type that detects a picture type in a preceding encoding process using the feature amount output from the DCT coefficient counter A detector, an encoding control unit that determines an encoding parameter at the time of re-encoding according to a detection result of the picture type detector, and re-encoding using the encoding parameter determined by the encoding control unit. And a coding unit that performs coding processing, so that when re-encoding an image signal that has undergone coding processing, the picture type can be detected from the image signal, thereby improving coding efficiency. The effect of reducing the processing circuit and the processing amount of the picture type detector can be obtained as compared with the coding apparatus.

The picture types detected by the picture type detector include intra-frame coded pictures (I pictures), forward inter-frame prediction coded pictures (P pictures), and bidirectional inter-frame prediction coded pictures (B pictures). One of the three picture types
By including one or more, it is possible to detect individual picture types for a decoded image signal group composed of I and P pictures and a decoded image signal group composed of three types of I, P, and B pictures. it can.

As a feature quantity counted by the DCT coefficient counter, a sum of absolute values or a sum of squares of DCT coefficients for each frequency domain is obtained.
By detecting the picture type in accordance with the time variation of the calculated sum of absolute values or sum of squares, it is possible to detect the picture type with a smaller processing amount as compared with a conventional re-encoding device.

Further, the picture type detector detects a picture in which the sum of absolute values or sum of squares in the high frequency region is smaller than the preceding and succeeding pictures as an intra-frame coded picture, thereby detecting an intra-frame coded picture. It is possible.

The picture type detector detects a picture in which the sum of absolute values or sum of squares in the low frequency region is larger than the preceding and succeeding pictures as an intra-frame coded picture or a forward inter-frame coded picture. It is possible to detect an inner coded picture and a forward inter-frame coded picture, in other words, it is possible to detect a bidirectional predictive coded picture.

Further, as the feature quantity counted by the DCT coefficient counter, the number of the DCT coefficients whose absolute value is larger than a predetermined threshold value or the number of those whose DCT coefficient values are smaller than the threshold value is obtained. By detecting the picture type in accordance with the obtained number, it is possible to detect an intra-frame coded picture.

The picture type detector detects the DC
It is possible to detect an intra-coded picture by detecting, as an intra-coded picture, a picture in which the absolute value of the T coefficient is smaller than the threshold or a picture having a smaller number than the threshold. is there.

Also, the encoding control unit determines the encoding parameter using the picture type detected by the picture type detector, thereby performing the encoding process with the same picture type as the preceding picture type. It is possible to perform efficient re-encoding processing with little image quality degradation.

Further, the encoding control unit determines the encoding parameter using the target generated code amount set in accordance with the picture type detected by the picture type detector, so that the same picture type as the preceding picture type is obtained. By performing the encoding process by type, it is possible to perform an efficient re-encoding process.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram illustrating a configuration of an image signal re-encoding device according to Embodiment 1 of the present invention.

FIG. 2 is an explanatory diagram showing frequency characteristics of 64 DCT counts output from a DCT unit 50 of FIG. 1;

FIG. 3 is an explanatory diagram showing power values of high frequency components (shaded portions in FIG. 3) of both horizontal and vertical components of DCT counting.

4 is a power value in which the horizontal component of the DCT coefficient is low and the vertical component is high (horizontal line in FIG. 3) or the DCT coefficient is high in the horizontal component and low in the vertical component (vertical line in FIG. 3). It is explanatory drawing which showed the electric power value of ().

FIG. 5 is an explanatory diagram showing a change for each picture of 64 DCT coefficients whose absolute value component is larger than a threshold value previously set in a DCT coefficient counter.

FIG. 6 is a block diagram showing a configuration of an image signal re-encoding device disclosed in JP-A-10-32829.

[Explanation of symbols]

11 encoding control unit, 12 encoding unit, 13 preprocessing unit, 50 DCT unit, 51 DCT coefficient counter, 52
Picture type detector.

Continued on the front page (72) Inventor Kazuo Sugimoto 2-3-2 Marunouchi, Chiyoda-ku, Tokyo Mitsubishi Electric Corporation (72) Inventor Fuminobu Ogawa 2-3-2 Marunouchi, Chiyoda-ku, Tokyo Mitsubishi Electric (72) Inventor Kotaro Asai 2-3-2 Marunouchi, Chiyoda-ku, Tokyo F-term (reference) 5C059 KK01 MA04 MA05 MA14 MA23 PP05 PP06 PP07 TA16 TA60 TC04 TC38 TD06 TD12 UA02 5J064 AA02 BA16 BB03 BC05 BC08 BC22 BD02

Claims (9)

[Claims] 1. An apparatus for re-encoding an image signal in which a decoded image signal that has undergone an encoding process is used as an input image signal and performs another encoding process, a DCT unit that performs a discrete cosine transform of the input image signal; Coefficient counter that counts a feature amount for each picture using a DCT coefficient output from a decoder, and a picture type detection that detects a picture type in a preceding encoding process using the feature amount output from the DCT coefficient counter A coding control unit for determining a coding parameter at the time of re-encoding according to a detection result of the picture type detector; and re-encoding using the coding parameter determined by the coding control unit. An image signal re-encoding device comprising an encoding unit for performing processing. 2. The image signal re-encoding device according to claim 1, wherein the picture type detector detects an intra-coded picture, a forward inter-frame predictive encoded picture, and a bidirectional frame as picture types to be detected. An image signal re-encoding device characterized by including at least two of three types of picture types of inter prediction encoded pictures. 3. The image signal re-encoding device according to claim 1, wherein the DCT coefficient counter includes D
Finding the absolute value sum or the square sum of each frequency domain of the CT coefficient, wherein the picture type detector detects the picture type according to the time variation of the determined absolute value sum or the square sum. Image signal re-encoding device. 4. The image signal re-encoding device according to claim 3, wherein the picture type detector performs intra-frame encoding on a picture in which a sum of absolute values or a sum of squares in a high-frequency region is smaller than the preceding and following pictures. An image signal re-encoding device characterized by detecting as a picture. 5. The image signal re-encoding device according to claim 3, wherein the picture type detector performs intra-frame encoding on a picture in which the sum of absolute values or sum of squares in a low-frequency region is larger than that of the preceding and following pictures. An image signal re-encoding device, wherein the image signal is detected as a picture or a forward inter-frame encoded picture. 6. The image signal re-encoding device according to claim 1, wherein the DCT coefficient counter has a feature value D
The absolute value of the CT coefficient is determined to be the number of those larger than a preset threshold or the number of those smaller than the threshold, and the picture type detector performs picture type detection according to the determined number. Characteristic image signal re-encoding device. 7. The image signal re-encoding device according to claim 6, wherein the picture type detector determines whether the number of pictures whose absolute value of the DCT coefficient is larger than a threshold is small or the number of pictures whose absolute value is smaller than the threshold is smaller than the threshold. An image signal re-encoding device characterized by detecting many pictures as intra-frame encoded pictures. 8. The image signal re-encoding device according to claim 1, wherein the encoding control unit determines an encoding parameter using a picture type detected by the picture type detector. Image signal re-encoding device. 9. The image signal re-encoding device according to claim 1, wherein the encoding control unit uses a target generated code amount set according to the picture type detected by the picture type detector. A re-encoding device for an image signal, wherein an encoding parameter is determined by performing an encoding process.